REVIEW PAPER

Glucose, insulin and other plasma metabolites shortlyafter injury

D. F. HEATH

North Western Injury Research Centre, Clinical Sciences Building, Hope Hospital, Salford

SUMMARY

A statistical study was made of measurements within3 h of injury on 533 patients grouped by injuryseverity using ISS. A scoring system was also usedthat took account of the number of injuries. Widelyaccepted hypotheses about the development ofhyperglycaemia were not supported. There was

evidence of inhibition of glucose metabolism ('insulinresistance'), but none for any particular mechanism.The only factor that was closely related within any

group to [glucose]* (plasma glucose concentration)was [lactate], the higher mean value and variance ofwhich after very severe injuries (ISS above 15)could account for much of the higher mean andvariance of glucose in this ISS range. At ISS 9-14,up to 1.4 h after injury, only 4% of the variance ofglucose could be accounted for by any combinationof the concentrations of catecholamines, cortisol,lactate and vasopressin, times of measurementafter injury and food intake, and injury severity andnumber of injuries. There was also no dependenceon the part of the body injured.

Injury increased the variability of [insulin] lessthan published statements imply. The increase foundwas entirely explicable, as expected, by the changesof [glucose] and [adrenaline].Key words: catecholamines, glucose, hormones,injury, ISS, insulin

INTRODUCTION

Correspondence:D.F. Heath,Northern WesternInjury ResearchCentre, ClinicalSciences Building,Hope Hospital, EcclesOld Road, Salford M68HD.

It is well known that plasma glucose concentration,[glucose], rises rapidly and very variably after injury.In brief, the accepted theory of the mechanism is asfollows. Hepatic output of glucose is increased bystimulation of hepatic glycogenolysis by adrenalineand glucagon and by gluconeogenesis from lactate,

* Concentrations in plasma are denoted by squarebrackets.

released in part from injured tissue and in part bystimulation of muscle glycogenolysis and glycolysisby adrenaline. It is possible that hepatic glycogen-olysis is also stimulated by vasopressin, angiotensin11 and activation of the sympathetic fibres in thehepatic nerve. Glucose uptake is inhibited by 'insulinresistance' involving cortisol and adrenaline. Theincrease of [glucose] with injury severity, at least asmeasured by ISS, is mediated entirely by the effectsof injury on [adrenaline].1 For reviews, see Barton2and Douglas & Shaw.3The above theory holds for the rat. The initial rise

in [glucose] can be accounted for by the actionof adrenaline, by gluconeogenesis from lactatereleased from damaged tissue and by adrenaline-stimulated glycolysis in muscle. There is no rise inglucose after 24 h starvation, when glycogen storesare depleted,4 insulin resistance develops rapidly5and depends upon increased [corticosterone].6The rise in [glucagon] probably contributes to themaintenance of hyperglycaemia,7 although the[glucagon]/[insulin] ratio is not raised.8

If this theory is correct then it should be able toaccount for the values of [glucose] that are measuredin patients following injury. Failure would indicatethat this theory is inadequate, however well it issupported by the results of controlled experimentsin animals.

Previous studies have not been promising9- 12 as

they rarely accounted for a quarter of the varianceof [glucose]. However, each study is open to criticismon one or more of the following counts. (1) Measure-ments made after treatment and/or several hoursafter injury were included in the analysis without fullinvestigation of the effects of treatment and time;(2) inappropriate statistical methods were used;(3) the interactions of the various factors were under-investigated; and (4) the results from patients with a

very wide range of severity of injury were lumpedtogether for analysis. The last factor, as shown

below, had particularly confusing effects.

Journal ofAccident andEmergencyMedicine 199411, 67-77

I have attempted a more complete study, confiningattention to measurements made within 3 h of injury,using multivariate analysis, and including somefactors not previously considered (i.e. specificity ofinjury and unconsciousness). In conjunction withISS a new scale was used, ISS-AIS, which tookaccount of all the injuries. The aim was to explainthe variability of the values of [glucose] (Fig. 1) and[insulin] (Fig. 2), which were measured within thefairly narrow bands of injury severity shown. These,suggested by Copes et al.,13 correspond to minor(ISS 1-3), moderate (4-8), severe (9-14), verysevere (16-24) and life-threatening (25+) injuries.Success has been measured by the percentageof the variance within each group that can beaccounted for by the various factors.Two sets of data have been used. The first data

set is that shown in Figs 1-4, and consists of all theplasma metabolite concentrations measured in theNorth Western Injury Research Centre up to mid-1990 on normal and injured subjects with the except-ion of those from patients with injuries with a knowntendancy to produce aberrant values, namely adultswith burns,14 children14 and those suffering from an

illness at the time of injury." This set correspondedto those used in most earlier studies, and had tobe used if comparisons were to made with thosestudies. However, this data set was statisticallyheterogeneous, and a subset in the age range17-65, with other limitations as described below,was used for more exact analysis.

Multivariate analysis had to be used with care asstatistical tests based on it require residuals (theportions of the measurements not accounted for) tobe normally distributed. It soon became clear thatthis was rarely the case. It is usual to neglect non-

normality, on the grounds that the method is 'fairlyrobust', i.e., the results are little affected by it.Robustness can, however, only be assessed byusing simulated data which may or may not reflectthe quirkiness of actual measurements. In the caseof the data used in this study non-normality mattered,because when residuals were made more normalby transforming the data to logarithms or squareroots, the percentage of the variance which wasaccounted for by theory was always increased,sometimes substantially. Because the aim was toaccount for as much of the variance as possible

(98)

(25)S 0

Normal 1 -3 4-8

Glucose

(288) (77)

(45)

0

*

Fig. 1. The distribution of plasmaglucose concentrations in normalsubjects and 533 patients within 3 hof injury. The bars representmedian values, the dashed line theupper 95% limit of the normalrange. The number in each group isshown in parentheses.

9-14 16-24 25+.l55

(93) (259)

(32) (24)(22) :

r/1

(72)

(41) 0

,FedFasted 1-3 4-8 9-14 16-24 25+

Normal ISSInsulin

Fig. 2. The distribution of plasmainsulin concentrations in normalsubjects and 489 patients within 3 hof injury. The bars representmedian values; the number in eachgroup is shown in parentheses.'Fed' refers to normal subjectsstudied 2-5 h after food.

D.F. Heath

20-

15-

1 5

EE 10-

5-

2-

128 -

64-

32 -

16-E

8-

4-

2-

1 -

68

.

Glucose andinsulin shortly afterinjury

19.2]

4.8]

E IE 1.2]

0.3 -

2.0

1.0

E

O J

100]

_ 10]

E 1.0-

0.1 -

100-

_10-

0E 1.0 -

0.1 -

401

, 20]

0-J

cN

Nl t_

LO }V

Lacatate

4s

NEFA cmaa i u

cm4~I0L

Adrenaline

N

-L-Noradenaline

CO

oN0

C

CO

CV) ~ ~ ~ ~ C

oI:dO4 g~~~c

Vasopressin

0

Aqo.-rea -asted Iss

NormalC c

X It +

- t a~~w e N

Fig. 3. The concentrations of plasma lactate, NEFA,adrenaline, noradrenaline and vasopressin measured innormal subjects and patients within 3 h of injury. The upperand lower limits of each box denote the central two thirdsof values, the whiskers the range and the bars themedians. Group sizes are shown. Individual values areshown where numbers are very low. 'Fed' refers tosubjects 2-5 h after food.

only the results with the 'best' (most normalizing)transformations are shown, along with the trans-formations themselves.

SUBJECTS AND METHODS

Subjects

There were 533 patients in the main set describedpreviously and 303 in the subset aged between 17and 65 years. Time after injury was usually knownto within 5-15min, and was recorded to 0.1 h.

69 Comparisons were made with healthy, normal

subjects aged 17-59 years. Most of the data havebeen published;1' 11'14-16 but the analysis is entirelynew. All work had received approval from the Salford(UK) Area Ethical Committee.

Injury severity

Severity was measured by the 1980 version of AIS,except for brain injuries, for which the 1985 versionwas used. For most other injuries suffered (fewwere penetrating) both versions gave the samescores.17 Relevant differences are noted.

ISS only takes account of the worst injury in eachof the three worst injured body areas. Two AIS-based systems that took account of all the injurieswere tested. The better, ISS-AIS, is used here, andconsists of ISS plus the sum of the scores not usedin its calculation. One patient, for example, with leginjuries, had 4 of AIS 3 to one leg and 4 of AIS 2 tothe other. The ISS was 9, the square of the highestAIS, but the ISS-AIS was 26, i.e., 9 + 9 + 8-namely the ISS plus the sum of the three unusedscores of 3 and the 4 of 2.

Blood collection and measurement of metabolites

A specimen of each patient's blood was taken formeasurement of metabolites shortly after arrival inhospital and usually before treatment, although 38patients had received analgaesics and 16 Haemac-cel. When more than one blood specimen wastaken, measurements from the first specimen onlyhave been included in this study. From normalsubjects blood was taken after an overnight fast of10-15h (fasted) or 2-5h after breakfast or lunch(fed).

At least four compounds were measured oneach blood specimen, by methods described in theoriginal papers except for [glucagon], that wasmeasured by a kit (Novo-Nordisk a/s, Bagsvaerd,Denmark), and vasopressin, that was measured byDr M. Forsling, Guys Hospital, London, UK. Themethod for measuring [cortisol] was changed in1978, and only later values have been included.The four values of [adrenaline] that were too low tomeasure were assigned the value 0.1 nmole I-1,which was close to the sensitivity of the method.

Data handling and statistical analysis

Searches and most of the statistics were carried outusing a purpose-written database on a PDP 11/34minicomputer. Some multivariate analysis was

coCo

0)

co~~~~~~~v

Cvf)

M FLNraNormal

Cv) C'

"1-

Fig. 4. Plasma cortisolconcentrations in normal subjectsand patients within 3 h of injurysubdivided by sex. Conventions areas in Fig. 3.

a)0c

M F M F M F M F M FL-Vj L--J L-V-J L-~J L-

1-3 4-8 9-14 16-24 25+I ~viSS

Cortisol

carried out using GLIM (General Linear InteractiveModelling package, Royal Statistical Society).

After extraction from the database, each set ofmeasurements was tested routinely for skewnessand kurtosis18'19 assuming normal, log-normal andsquare-root-normal distributions. Because of theprevalence of non-normality, simple comparisonsand correlations were carried out using the Wilcoxonrank sum test and Kendall's rank correlation test.These were shown to be more efficient thanparametric methods (t-test and simple regression),as expected20 for such non-normal data.

Multivariate regression was first carried out usingthe three or four independent variables that wereexpected to account for most of the variance of thedependent variable either from the results of sim-ple regression or on pharmacological grounds -[adrenaline], for instance, was always included. Theother candidate variables were then added one at atime, and retained if the residual variance wasdecreased by more than 2%. Finally, variables fromthe remaining set were removed one at a time (step-down method) to obtain the best predictors of thedependent variable. An independent variable is saidto be without effect if it accounted for less than 2%of the variance of the dependent variable.

RESULTS

Preliminary observations

Non-normality. Of the 25 sets of data large enoughto test (N > 24) 10 were skew or kurtotic (P < 0.05)on all three distributions tested, normal, log-normaland V-normal.

Analgaesics, Haemaccel and unconsciousness.Plasma hormone and metabolite concentrations inpatients were unaffected by analgaesics and/orHaemaccel (tested on the largest group, ISS 9-14)or by unconsciousness. The last was tested bycomparing the 14 patients in the subset agedfrom 17 to 65 years at ISS 25+ who were uncon-scious for at least 1 h with the 22 who were not.There were no significant (P < 0.05) differences.Pain and awareness had, apparently, little effect.Normal subjects. When values were affected by

fasting or sex they are shown separately. [Glucose]was not significantly affected by either. Catechol-amines, glucagon and vasopressin were onlymeasured in subjects 2-5 h after food (fed).

Hyperglycaemia. Many patients were not hyper-glycaemic: half of those up to ISS 8, were within thecentral 95% of the normal range, as were over a

quarter of patient with ISS of 9-14 and a few athigher ISS. Values did not peak at 2 h after injury, asthey tend to do in children with burns.14 Hyper-glycaemia at ISS 1-3 was probably transient,because all values from 1.5h were normal; and[glucose] rose with time at ISS 25+ (P<0.005).Otherwise there was no time-dependence. Medianvalues of [glucose] have been shown elsewhere toremain constant in each band from ISS 4-8 upwardsfor at least 8 h.'4

[Insulin]. Every set (Fig. 2) was roughly log-normally distributed. Median values after injury wereclose to those in normal non-fasted subjects andhigher than in fasted normal subjects (P <0.001except at ISS 25+, when P< 0.05). There was no

dependence on time after injury.[Lactate]. Values tended to be much higher at

D.F. Heath 2.0 -

1.5 -

E1.0 -

0 5 -

0

70

ISS 16+ (Fig. 3). They tended to fall with time, andat ISS 9-14 a far smaller porportion (18/103)were above normal from 1.5 to 3.0 h than earlier(68/147).[NEFA]. Except at ISS 25+ values (Fig. 3) tended

to be higher than in normal fed subjects (P < 0.001)and similar to those in normal fasted subjects. AtISS 25+ values were intermediate. They tended torise with time, but only significantly at ISS 9-14(P < 0.01).[Catecholamines]. Values (Fig. 3) tended to rise

with ISS, as reported elsewhere,' 10 and to fall withtime, but only significantly at ISS 9-14 (P <0.05).

[VasopressinJ. Values were greater and morevariable after injury (Fig. 3). Most were obtained atISS 9-14.[Glucagon]. The five values after injury were all

measured within 1 h and at ISS 16-42. The valueswere a little higher than in normal subjects andsimilar to others published.21

[CortisolJ. Values (Fig. 4) were higher in womenthan in men at ISS 4-8 (P<0.05) and 9-14(P < 0.005), as expected,11 and rose with time afterinjury at ISS 4-8, 9-14 and 25+, significantly atISS 9-14 (P<0.005 in both men and women).

Simple non-parametric correlations

Theory and published correlations over wide rangesof ISS lead one to expect positive correlationsbetween [glucose] and [adrenaline], [noradrenaline]and [vasopressin] and between [NEFA] and [nor-adrenaline] and negative correlations between[glucose] and [NEFA] and between [insulin] and[adrenaline]. No such correlations were found withinany of the narrower ISS bands used here. [Glucose]rose with [cortisol] in females (P < 0.01 at ISS 4-8and 9-14) but not in males. Age had little effect atany ISS on any metabolite. [Glucose] and [lactate]were correlated weakly with injury severity in anyISS range, more so with ISS-AIS than with ISS,though rarely significantly.

Time after food. Time after food had been recordedin 189 patients, in 26 of whom it was in the range10-20 h (median 14 h). As expected, [NEFA] tendedto rise with time after food, and [lactate], [glucose]and [insulin] to fall, but the correlation coefficientswere very small, particularly so with [glucose].

Multivariate analysis

Glucose. The most normalizing transform for[glucose] was /[glucose]. The independentvariables were times after injury and food, ISS-AIS,

log[adrenaline], log[noradrenaline], log[cortisol],log[lactate] and [vasopressin]. [Insulin] was nottreated as an independent variable, because it wasalways positively related to [glucose], i.e., [glucose]controlled [insulin], not [insulin] [glucose]. Therewere too few values of [glucagon] to include.

Initially, the data used were those from the 78patients in whom [glucose], [lactate] and [adrenaline]had been measured. To increase sample sizes,they were divided into only two ISS ranges, 1-14

and 16+ (Table 1, first two columns).The greatest dependence was on log[lactate],

more at ISS 16+ than at ISS 1-14. When log[lac-tate] was a variable there was no dependence on

ISS-AIS or log[adrenaline], nor on any other par-ameter. In the absence of log[lactate] there was

some, considerably smaller, dependence on ISS-AIS and log[adrenaline]; but the latter had no effectsindependently of the former. Because [adrenaline]tended to fall with time after injury, the analysis wasrepeated using only data from the first hour; but thedependence on [adrenaline] was not higher, but(insignificantly) less. With log[adrenaline] and [vaso-pressin] as the only independent variables at ISS1-14 the dependence on [vasopressin] was

(not significantly) negative, as it was in simpleregression.Because the effects of [adrenaline] could be sub-

sumed entirely under those of [lactate], an investi-gation was attempted on all patients whether or

not [adrenaline] had been measured in them. Theresiduals were, however, very non-normally dis-tributed. This problem was overcome by using thesubset aged between 17 and 65 years.

Table 1. Percentage reduction of variance of V[glucose]due to regression in patients with measured [adrenaline]

All Subset*ISS range 1-14 16+ 1-14

N 46 32 24ParametersTAlt 0 2 24ISS-AIS 10 19 1

LogADR 8 13 23LogLAC 27 45 2LogLAC + TAI 27 48 11ISS-AIS + logADR 10 20 12

LogLAC + ISS-AIS 30 46 24LogLAC + LogADR 28 45 12LogLAC + logADR + 30 47 24ISS-AIS

* Subset aged 17-65 yearst Abbreviations: TAI, time after injury; ADR, [adrenaline];LAC, [lactate].

Glucose andinsulin shortly afterinjury

71

Members of this group were also without evidenceof ethanol intake and survived at least 24 h. Age16and ethanol1" have known affects on [glucose] and/or [lactate]. The effects of imminence of death wereshown by comparing the 16 patients at ISS 25+who died within 4 h with the 56 who survived at least24 h. (Only four lived 4.1 -24 h.) In the 16 patients atISS 25+ the values of [lactate] (P < 0.001), [insulin](P < 0.01) and [noradrenaline] (P < 0.01) weregreater, those of [NEFA] (P < 0.05) and [cortisol](P < 0.001) were less, and those of [glucose]were more variable, 1.8- 19.0 against 4.2-14.8 mmole 1. The 24-h limit was used in order toleave a safety margin.Some results of multivariate analysis are shown

in Table 2. Those at ISS 1-3 are excluded, becausevalues of [glucose] were all normal by 1.5 h, as are amass of near-zero correlations, such as those withsex, time after food and [cortisol]. Time after injurywas often a factor, most markedly at ISS 9- 14, andfor this group the results up to and from 1.5 h areshown separately.

At ISS 4-8 ISS-AIS was the only significantfactor. At ISS 9-14 up to 1.4 h no combination ofcatecholamine, cortisol and lactate concentrations,ISS-AIS and time after injury could account for morethan 4% of the variance (combination not shown inTable 2); and neither [vasopressin] nor time afterfood could account for as much as 2%. From 1.5 hat this ISS, however, ISS-AIS and [lactate] hadsignificant and partly independent effects. At ISS16-24 [lactate] was the only factor, whereas at ISS25+ time after injury was equally important; but inboth ranges injury severity (ISS-AIS) was a negligiblefactor. With log[lactate] as a factor there was nodependence on [cortisol] or time after food.There are some discrepancies between Tables 1

and 2. The different dependency on [lactate] largely

disappeared when the analysis of patients in whom[adrenaline] had been measured was confined tothose who were in the subset aged between 17 and65 years (Table 1, last column), thus excluding someelderly and one drunk patient. The higher depen-dence on ISS-AIS in Table 1 was wholly attributableto two high values of [glucose] in patients with highISS-AIS from multiple limb injuries which had agreater effect on the statistics of the smaller group.Some, but not all, of the greater dependence ontime after injury at ISS 25+ shown in Table 2 than atISS 16+ shown in Table 1 can be explained by thecombining of the two sets, ISS 16-24 and ISS 25+,in Table 1, and by a slight mismatch of times afterinjury.

In summary, few predictors of [glucose], evenwhen significant, could account for more than a fewper cent of the [glucose] variance. The main excep-tions were [lactate] at ISS 16+ and time after injuryat ISS 25+. There was negligible dependence ontime after food. No combination of factors couldexplain more than about 4% of the variance at ISS9-14 up to 1.4h.

Insulin. The main group, unselected by age, etc,was used. [Insulin] depended only on [glucose] and[adrenaline], both regression coefficients beingstronger in multiple than in single regression, owingto the weak positive correlation of values of [glucose]and [adrenaline] (Table 3).The overall multiple regression coefficient for

log[insulin] against V[glucose] was about 0.5 + 0.1which is very similar to that (0.53) calculable fromthe results of a hyperglycaemic clamp after an over-night fast22 in which a rise of [glucose] from 5.3 to12.2 mmole I- 1 raised insulin from 10 to 43 mU - 1.Thus the insulin response to the rise in glucose wasessentially normal when allowance was made forthe counter-action of adrenaline.

ISS range 4-8 9-14 16-24 25+TAI (h)- 0-1.4 1.5-3.0

N 59 89 57 22 32ParametersTAI 0 1 0 3 20ISS-AIS 14 3 10 3 1LogLAC 0 2 12 51 22LogLAC + TAI 1 2 14 57 41LogLAC + ISS-AIS 16 4 17 59 23Significantt ISS-AIS None LAC LAC LAC,TAI

* Abbreviations are as in Table 1.t Coefficients of regression on these parameters or their logarithms weresignificantly (P < 0.05) different from zero in multiple regression.

Table 2. Percentage of variance of/[glucose] due to regression in the

subset aged 17-65 years

D.F. Heath

72

Because the effects of injury on [insulin] were

mediated by the changes in [glucose] and [adrena-line], and these changes accounted for about a thirdof the variance of log[insulin] after injury, injuryshould have increased the variance of log[insulin]by about 50%. The data yielded three estimates ofthe variance ratio, injured/normal, that were con-

sistent with this view. Because most patients hadnot been without food for long, the first estimate isthat from all the 489 patients in whom [insulin] hadbeen measured with the 32 normal subjects 2-5 hafter food. The ratio was 1.46, i.e., injury increasedthe variance by 46%. The values of the ratio whenpatients and normal subjects were matched for timeafter food were 1.49 (2-5h) and 1.79 ( 10h).NEFA. Values of [NEFA] were expected to be

negatively related to those of [lactate], and, by theglucose-fatty acid cycle, to [glucose] and [insulin],and positively related to [adrenaline] and [nor-adrenaline] owing to the lipolytic action of cat-echolamines and sympathetic nerve stimulation(reviewed2). Neither simple nor multiple regressionrevealed any such relationships. The only significantrelationships were against expectation, negativewith log[noradrenaline] at ISS 16+ (P < 0.01), andpositive with log[insulin] at ISS 9-14 (P < 0.05).

Effects of particular injuries

Some of the variance of [glucose] might have beenexplained if there had been injury-specific effects on[glucose]. However, this was not the case.

From the subset aged from 17 to 65 years, 101patients were selected whose injuries, except forminor abrasions and contusions, were confinedto a single body area. Upper and lower limbs were

regarded as separate areas, unlike in AIS. The ISS

range was 9-1 1. The group with lower limb injurieswas further subdivided into those with injuries con-

fined to hip, thigh, knee, tibia and fibula, or ankle.Values of [glucose] are shown in Fig. 5. For each

compound each subset was compared with theoriginal group minus the subset, e.g., 11 patientswith fractured femur were compared with the 90patients with other injuries. There were 40 compari-sons involving [glucose], [lactate], [NEFA], [insulin]and [cortisol] with sample sizes large enough to test.Two were 'significant' (0.05 < P < 0.01), the numbermost likely to occur by chance on the null hypo-thesis. There were too few measurements of othercompounds to test. Interestingly, the effects of upperlimb injuries, which have been downgraded to AIS 2in versions of AIS later than that used (1980), were

indistinguishable from those of other injuries stillscored 3.

DISCUSSION

Values of [insulin] are easier to explain than those of[glucose]. Of the increase in variance of log[insulin]after injury, about 50%, was wholly explicable bythe increased variability of [glucose] and [adrena-line]. Mean values were similar to those in normalsubjects 2-5h after food. They did not vary withtime after injury.These findings are contrary to the usual view,

based mostly on results given by Vitek et al.12according to which [insulin] is raised by all but life-threatening injuries, is very much more variablethan normal, and tends to rise with time. This analy-sis was somewhat distorted by the use of the normaldistribution for their highly skewed data and, becauseit is unlikely that many of their patients were fasted,by the use of fasted normal subjects as controls.

Table 3. Dependence of log[insulin] on V[glucose] and log[adrenaline].

ISS range (N) 1-14 (29) 16+ (26) All (55)Simple regression

%var coefft %var coefft %var coefft

LogADR 2 -0.03 ± 0.05 0 -0.01 ± 0.04 1 -0.02 ± 0.03VGLU 28 0.64±0.19 29 0.39±0.12 17 0.32±0.09Multiple regressionLogADR 35* -0.09± 0.04 39* -0.08± 0.03 34* -0.10± 0.03VGLU 0.75+0.18 0.52 ± 0.12 0.53± 0.10

Abbreviations: %var, percentage reduction in variance due to regression; coefft, regressioncoefficient (simple or multiple) ± SEM; ADR, [adrenaline]; GLU, [glucose].* Total reduction due to multiple regression.73

Glucose andinsulin shortly afterinjury

* .

0* * *

2 2* em _ 0 0

0 0

2 3

20*02

*o se*

.

0

2

0ZZ0e ed

m0 0

3 4 5 6 7 8 9

[Glucose] mmol L-1

2

10

Fig. 5. Comparison of plasma glucose conc

group of patients and in subgroups with pa

drawn from it. The groups are further descritText. The number in each group is shown

Recalculation of my data on their assui

results very similar to those on that gpatients who were studied within 3.1 tarrival at hospital. They, however,another group of patients, referred frorrtals, 3-6 h after injury. These patientstressed and most had received treatrrvalues of [insulin] were higher with %

times as great. The impression of r

variance and time-dependence comesing the two groups, although Vitekcareful to distinguish them.

Frayn et al.1 related log[glucose], lo(

log[adrenaline] in 40 patients, three o

excluded from my study. They found a

tion between the 20 patients in whornexceeded 2 nmole 1 -1 and in 20 patierwas less - all six values of [insulin] abcwere in the patients with low [adrenalinsuch difference in my larger group.

(Figs 2 & 3) confirm this view. Thus, tiof patients with [insulin] greater than 2almost the same at ISS 4-8 (25/11[adrenaline] rarely exceeded 2 nmole25+ (14/72), at which it was rarelyet aW.'s result may well have been a chbecause the probability that one of twosubjects will have no [insulin] greater this 0.03 [calculated from the binomial theto the bulked proportion (25 + 14)/(1 1

Causes of hyperglycaemia

At ISS 1-3 hyperglycaemia was only tthis group is not considered further.

Chest (3) The two main findings are the almost completefailure to account for the variability of [glucose] up to

Head (5) ISS 14, especially in the subset aged between 17

Upper limbs (19) and 65 years, and the swamping effects of the rise

Pelvis (3) in [lactate] after more severe injuries. In this group

Thigh (11) (ISS 16+) the mean value of -[glucose] and its

Hip (9) variance were higher than at ISS 4-14, and 65% ofthe difference in the mean value and 85% of the

Tib. and Fib. (47)difference in variance could be accounted for by the

All (101) differences in the values of [lactate]. (Calculationwas from the regression equations for V[glucose]on log[lactate] in the two groups).

Thus, hyperglycaemia had two components, a

centrations in a [lactate]-independent one at all ISS, superposed onrticular injuries which was a [lactate]-dependent component, whichbed in the Main was almost confined to patients with very severe orn parentheses. life-threatening injuries.

This picture is very unlike that given by earliermptions gave studies, referred to in the Introduction, in which3roup of their [glucose], [lactate], and [adrenaline] were positivelyh of injury on correlated over the whole range of ISS. Inspectionalso studied of Figs 1 to 4, however, shows that these correlationsn other hospi- merely reflect the facts that the median values ofts were more [glucose], [lactate] and [adrenaline] were all highernent and their at ISS 16+, so that combining this group with thosevariance four at lower ISS was bound to create positive corre-

nuch greater lations between all three concentrations and of eachfrom combin- with ISS even though the correlations within the ISSet al. 12 were 4-14 group were very poor.

The [lactate]-independent and [lactate]-dependentg[insulin] and components of [glucose] variability are now con-

4 whom were sidered in turn.clear distinc-

i [adrenaline] The flactate]-independent componentits in whom itve 20 mU 1 For brevity only the results at ISS 9-14 are

e]. I found no discussed.Other results Accepted theory failed comprehensively tohe proportion account for this component of variance, especially!o mU 1- 1 was in the first 1.4 h. Failure to demonstrate a mechanism17), at which does not, of course, establish that it does notl-1, as at ISS operate, and there were a number of factors thatless. Frayn could not be investigated properly. These are con-

lance finding, sidered briefly below, but it is hard to avoid theigroups of 20 conclusion that there are features of the hyper-ian 20mU I -1 glycaemic response in humans that are not under-borem applied stood and that are not covered by the mechanisms7 + 72)]. proposed so far.

Variability of hepatic glycogen reserves. Timeafter food accounted for under 2% of the variance of/[glucose]. This finding at first seems incompatible

transient, and with the almost complete suppression of hyper-glycaemia in the rat by 24 h of starvation.4 This

D.F Heath

74

Glucose and period in the rat, however, is probably equivalent toinsulin shortly after at least 48 h of fasting in humans, and none of theinjury patients had fasted for more than 20 h. Liver glyco-

gen has been measured after an overnight fast,23and was, on average, equivalent to a plasma glucoseconcentration of 30 mmol - (assuming a glucosedistribution volume of 250mlkg-1), much higherthan any measured. Thus, glycogen reserves wereunlikely to have been a limiting factor in manypatients.

Variability in the sympathetic stimulation of thehepatic nerve. Because hepatic nerve stimulationcannot be measured directly in patients the onlyevidence for sympathetic activity is the rise in [nor-adrenaline], which was always found, and whichwas considerable after more severe injuries. Corre-lations between [glucose] and [noradrenaline] were,however, all trivial. As is well known, plasma nor-adrenaline is a poor index of sympathetic stimu-lation, because it merely represents spill-over fromsynapses. A rise in [noradrenaline] is, however,good evidence of a rise in sympathetic stimulationsomewhere in the body, so that, for sympatheticstimulation to account for much of the variation of[glucose], it must be assumed that the degree ofsympathetic stimulation to the hepatic nerve wasalmost completely unrelated to that elsewhere,and was, by implication, highly specific. If so, whatpurpose did it fulfil? It plainly failed to producehyperglycaemia in many recently fed patients. Thehypothesis seems to be too complex to be plausibleand impossible to investigate, however, it cannot becompletely ruled out.

Variations in the concentrations of hormones thatwere not measured. There were no measurementsof [angiotensin-l1] and few of [glucagon]. There issome indirect evidence from studies after burn injury,which causes more hyperglycaemia than the injuriesconsidered here, that angiotensin-ll is an unlikelycandidate (see Barton2 for references). The signifi-cance of the fairly slow rise of [glucagon] afterinjury21 has been much debated.2 It is less than inthe rat,8 but is generally held to stimulate output ofglucose from the liver after the first hour. Moremeasurements are needed (see below).The rise in [adrenaline] was a poor index of the

stimulation of glucose output by adrenaline.Because [adrenaline] fell with time, an initially closerelationship between [glucose] and [adrenaline]might have been lost by the time the blood speci-mens were taken. The evidence is against thishypothesis. On it the [glucose] - [adrenaline]

injury, whereas it was slightly better 1.5-3h afterinjury than before. Moreover, after more severe

injuries, when [adrenaline] was sometimes very

high, the relationship was still poor.

Variable inhibition of peripheral glucose metab-olism -'insulin resistance'. There is marked insulinresistance in the rat shortly after both dorsal scaldand hind-limb ischaemia5,24 and also after injury inother species.25 In humans there is variable inhi-bition of glucose oxidation within 3h of injury.26 Inthe present study the values of [NEFA] impliedconsiderable inhibition of glucose uptake and/ormetabolism. Thus, [NEFA] was at fasting levels(except at ISS 25+, when impairment of circulationprobably reduced output) even though both [glu-cose] and [insulin] tended to be above normal fasted,and [NEFA] was not negatively correlated with either[glucose] or [insulin]. The alternative explanation,that [NEFA] was raised by sympathetic stimulationof lipolysis, is unlikely to be true, because it wasnegatively, not positively correlated with [noradrena-line] even in multilinear regression with [glucose] as

a second independent variable.In principle, variations in inhibition of peripheral

glucose metabolism and/or uptake could accountfor much of the variance of [glucose].25 The mechan-ism of inhibition is, however, far from clear. Thatwhich is usually postulated, namely synergismbetween adrenaline, cortisol and glucagon, is basedon infusion experiments on healthy adults,22'27'28but the requisite concentrations of all three hor-mones are higher than those in many patients.Thus, up to ISS 14 [adrenaline] was rarely highenough (about 2 nmole I- 1), [cortisol] was too low inmany men and some women, and [glucagon] was

probably also too low. After more severe injuries[cortisol] was rarely high enough in either men or

women. It is also not clear how ever-changing syner-

gistic effects could cause [glucose] to stay nearlyconstant for several hours.14A convincing test would require many more

measurements on individual patients than havebeen carried out so far. A feasible protocol wouldinvolve sequential measurements on a group ofbroadly similar patients with ISS 9-14, e.g. young

men with multiple lower limb injuries. The first bloodspecimen would be taken within 30 min of injury,and others at intervals not exceeding 30 min for atleast 21/2h. Compounds measured would include

glucose, lactate, insulin, cortisol, glucagon, adrena-

line, noradrenaline, vasopressin and angiotensin I1.Blood gases would be measured at least once,

75 relationship should have worsened with time after and an assessment of blood loss would be made.

D.F. Heath Nutritional status, bodily habit and previous fitnesswould be noted.Such a study should also reveal to what extent

[glucose] does remain constant from 30 min in indi-vidual patients. In the main, previous work hasshown constancy of group mean values.

The [lactate] - dependent component

The effects of [adrenaline] and injury severity on[glucose] were very much smaller than those of[lactate]; and most of the lactate-dependent com-ponent of [glucose] variability must be attributed tothe conversion of the variable amounts of lactate toglucose. The fact that the component was practicallyconfined to patients with very severe injuriessuggests that the controlling factor was hypoxia,regional from tissue damage or general from bloodloss, which increased glycolysis to lactate.The disappearance of the effects of injury severity

and [adrenaline] on [glucose] when the analysisincluded [lactate] does not imply that they did notexist. There are two patterns to be considered. Inone, effects are mediated via [lactate]. Thus bloodloss and hypoxia are likely to rise with injury severityand stimulate glycolysis, as will, in muscle, a rise in[adrenaline]. These effects on glycolysis will appearas changes in [lactate]. Because, in this mechanism,lactate is nearer to glucose production in the causalchain than ISS-AIS or adrenaline, the effects ofthese factors will be subsumed under those oflactate in the statistical analysis. In the other pattern,adrenaline and, from the effects of blood loss, etc.,vasopressin and sympathetic stimulation may stimu-late hepatic output of glucose directly, with effectsthat should be independent of [lactate]. Either theseeffects were very small in the patients studiedor their apparent elimination when [lactate] wasincluded was a statistical artefact. The latter cannotbe ruled out. All these factors act in the samedirection, with lactate having the greatest effect.Cases are known in least mean squares analysis inwhich the effects of a minor factor disappear when amajor one is introduced with effects in the samedirection.

Injury severity scores

Whether measured by ISS or ISS-AIS, injury severitywas never a good predictor of [glucose]. ISS-AISwas only marginally better. This was surprising,because ISS-AIS includes the AIS values of allthose injuries not included in the calculation of ISS,

and consequently gave a much more extendedscale (9-29 against 9-14 for example), with fewertied values. Therefore, it seems unlikely that anypurely anatomical description of injury severity couldever predict the metabolic response at all well.

Other injury scales based on severity of comaand estimates of impairment of circulation could notbe tested, but seem unlikely to do much better,because the metabolic response was not affectedby unconsciousness, and because the negligiblecorrelation of [glucose] with [lactate] during the first1.4 h at ISS 9-14 indicates that blood loss was nota major factor at this severity.

GENERAL COMMENTS

It is apparent that there are still great gaps in ourunderstanding of hyperglycaemia after injury.Many necessary measurements have not beenmade, and those that have been do not provide anadequate explanation.The matter may be of more than academic

interest. The debilitating loss of protein that takesplace 1 to 2 weeks following injury is known to berelated to reduced insulin sensitivity.29 Is there anyrelationship between this reduction and the earlyreduction? On the present analysis, these should berelated, up to ISS 14, to the degree of hypergly-caemia. As far as I know this has never beeninvestigated.

ACKNOWLEDGEMENTS

I thank all those colleagues who have allowed me tomake free use of their data, even when my inter-pretations have conflicted with theirs, and especiallyJohn Rose, who carried out the glucagon measure-ments for this work, and the Salford (U.K.) AreaHealth Authority for financial support.

REFERENCES

1. Frayn K.N., Little R.A., Maycock P.F. & Stoner H.B.(1985) The relationship of plasma catecholamines toacute metabolic and hormonal responses to injury inman. Circulatory Shock 16, 229-240.

2. Barton R.N. (1985) Neuroendocrine mobilization ofbody fuels after injury. British Medical Bulletin 41,218-225.

3. Douglas R.G. & Shaw J.H.F. (1989) Metabolicresponse to sepsis and trauma. British Journal ofSurgery 76, 115-124.

4. Stoner H.B. (1958) Studies on the mechanism of76

Glucose andinsulin shortly afterinjury

shock. The quantitative aspects of glycogen metabol-ism after limb ischaemia in the rat. British Journal ofExperimental Pathology 39, 635-651.

5. Frayn K.N. (1975) Effects of burn injury on insulinsecretion and on sensitivity to insulin in the rat in vivo.European Journal of Clinical Investigation 5,331-337.

6. Barton R.N. & Passingham B.J. (1980) Evidence for arole of glucocorticoids in the development of insulinresistance after ischaemic limb injury in the rat.Journal of Endocrinology 86, 363-370.

7. Heath D.F. (1980) Carbohydrate metabolism afterinjury. The development and maintenance of hyper-glycaemia. Advances of Physiological Sciences 26,63-70.

8. Rose J.G. & Heath D.F. (1986) The effects of stressand injury on the activity of phosphoenolpyruvatecarboxykinase in the liver of the rat. BiochemicalJournal 233, 239-244.

9. Davies C.L., Newman R.J., Molyneux S.G. &Grahame-Smith D.G. (1984) The relationship betweenplasma catecholamines and severity of injury in man.Journal of Trauma 24, 99-105.

10. Kenney P.R., Allen-Rowlands C.F. & Gann D.S.(1983) Glucose and osmolality as predictors of injuryseverity. Journal of Trauma 23, 712-719.

11. Stoner H.B., Frayn K.N., Barton R.N., Threlfall C.J. &Little R.A. (1979) The relationships between plasmasubstrates and hormones and the severity of injury in277 recently injured patients. Clinical Science 56,563-573.

12. Vitek V., Lang D.J. & Cowley R.A. (1979) Admissionserum insulin and glucose levels in 247 accidentvictims. Clinica Chimica Acta 95, 93-104.

13. Copes W.S., Champion H.R., Sacco W.J., LawnickM.M., Keast S.L. & Bain L.W. (1988) The injury severityscore revisited. Journal of Trauma 28, 69-77.

14. Childs C., Heath D.F., Little R.A. & Brotherston M.(1990) Glucose metabolism in children during the firstday after burn injury. Archives of Emergency Medicine7, 135-146.

15. Barton R.N., Stoner H.B. & Watson S.M. (1987)Relationships among plasma cortisol, adrenocortico-trophin, and severity of injury in recently injuredpatients. Journal of Trauma 27, 384-392.

16. Frayn K.N., Stoner H.B., Barton R.N., Heath D.F. &Galasko C.S.B. (1983) Persistence of high glucose,insulin and cortisol concentrations in elderly patients

with proximal femoral fracture. Age and Ageing 12,70-76.

17. Copes W.S., Lawnick M.M., Champion H.R. & SaccoW.J. (1988) A comparison of abbreviated injury scale1980 and 1985 versions. Journal of Trauma 28,78-86.

18. Geary R.C. (1936) Moments of the ratio of the meandeviation to the standard deviation for normal samples.Biometrika 25, 295-305.

19. Snedecor G.W. & Cochran W.G. (1986) StatisticalMethods, 7th ed. Iowa State University Press Ames.

20. Bradley J.V. (1968) Distribution-free Statistical Tests.Prentice-Hall International Inc, London.

21. Meguid M.M., Brennan M.F., Aoki T.T., Muller W.A.,Ball M.R. & Moore F.D. (1974) Hormone-substrateinterrelationships following trauma. Archives ofSurgery 109, 776-783.

22. Black P.R., Brooks D.C., Bessey P.Q., Wolfe R.R. &Wilmore D.W. (1982) Mechanism of insulin resistancefollowing injury. Annals of Surgery 196, 420-435.

23. Nilsson L.H:son (1973) Liver glycogen content in manin the postabsorptive state. Scandinavian Journal ofClinical and Laboratory Investigation 32, 317-323.

24. Heath D.F. & Corney P.L. (1973) The effects of star-vation, environmental temperature and injury on therate of disposal of glucose by the rat. BiochemicalJournal 136, 519-530.

25. Heath D.F. (1986) Experimental studies on energymetabolism after injury and during sepsis. In: LittleR.A. & Frayn K.N. (eds) The Scientific Basis for theCare of the Critically III, pp 95-96. Manchester Univer-sity Press, Manchester.

26. Little R.A., Stoner H.B. & Frayn K.N. (1981) Substrateoxidation shortly after accidental injury in man. ClinicalScience 61, 789-791.

27. Gelfand R.A., Matthews D.E., Bier D.M. & SherwinR.S. (1984) Role of counter-regulatory hormones inthe catabolic response to stress. Journal of ClinicalInvestigation 74, 2238-2248.

28. Shamoon H., Hendler R. & Sherwin R.S. (1981) Syner-gistic reactions amongst antiinsulin hormones in thepathogenesis of stress hyperglycemia in humans.Journal of Clinical and Endocrinology and Metabolism52, 1235-1241.

29. Frayn K.N. (1986) Hormonal control of metabolism insepsis and trauma. Clinical Endocrinology 24,577-599.

77